|
| 超像素级局部对比的声图水下小目标检测方法 |
| Superpixel-level local contrast method for underwater small target detection in sonar images |
| 投稿时间：2024-09-25 修订日期：2025-04-25 |
| 中文摘要: |
| 针对声纳图像信噪比低和样本少等引起的水下小目标检测率低和虚警高的问题，提出了一种超像素级局部对比的水下小目标检测方法。该方法利用了简单线性迭代聚类算法，将相似强度的相邻像素自适应分组，构造超像素声图；利用声图增强和分割方法，通过局部超像素分组、剔除平均统计和浓度比局部对比增强等处理，有针对性地增强目标，抑制复杂背景，进而提高目标检测率；结合声图信噪比、浓度及功率等统计参数，对声图感兴趣区域进行统计评估和筛选，降低虚警率。经过真实声纳图像验证，该方法能够有效提高小目标检测率和降低虚警率，尤其适用于样本少和信噪比低的水下小目标检测。 |
| 英文摘要: |
| A superpixel-level local contrast method of underwater small target detection is proposed to address the problems of low detection rate and high false alarms caused by low signal-to-noise ratio and few samples in sonar images.The proposed method employs the Simple Linear Iterative Clustering (SLIC) algorithm to adaptively group adjacent pixels with homogeneous intensity, thereby generating superpixel-structured sonar images. Through sonar image enhancement and segmentation techniques, the method achieves targeted enhancement of true targets and suppression of complex backgrounds via three key operations: local superpixel clustering, elimination of statistically averaged interference, and concentration-ratio-based local contrast enhancement, collectively improving target detection accuracy. Furthermore, statistical evaluation and region-of-interest (ROI) screening are implemented by integrating critical parameters including the signal-to-noise ratio (SNR), spatial concentration, and acoustic power characteristics of sonar images, effectively reducing false alarm rates. Experimental validation on real-world sonar datasets demonstrates that this methodology significantly enhances small target detection performance while maintaining low false alarms, particularly advantageous for underwater scenarios characterized by limited training samples and low-SNR conditions |
| DOI： |
| 中文关键词: 水下小目标检测 声纳图像 局部对比度 超像素级 |
| 英文关键词: Underwater small target detection Sonar images Local contrast Superpixel level |
| 基金项目: |
|
| 摘要点击次数: 37 |
| 全文下载次数: 0 |
|
查看/发表评论 下载PDF阅读器 |
| 关闭 |
